DataViz Makeover 2

DataViz Makeover

Dive deeper into survey results of public’s willingness on Covid-19 vaccination.

Bai Xinyue
02-19-2021

Data Visualisation Link (Tableau Public) - https://public.tableau.com/profile/xinyue.bai#!/vizhome/ofStronglyAgree-gettingCOVID/DatavizMakeover2?publish=yes

1. Background Information

For this visualisation makeover, I have used data from Imperial College London YouGov Covid 19 Behaviour Tracker Data Hub. This data gathers global insights on people’s behaviours in response to COVID-19 covering 29 countries, in the form of survey questionnaire. In particular, this post is interested in exploring the willingness of the public on COVID-vaccination. In this blog, I will makeover visualisation on vaccine willingness done by one of the research scientists, by examining the following 3 survey questions in the context of different gender and employment status:
1. If a Covid-19 vaccine were made available to me this week, I would definitely get it.
2. I am worried about getting COVID19.
3. I am worried about potential side effects of a COVID19 vaccine.

2. Critiques and Suggestions

original visualisation

Figure 1: original visualisation

2.1 Clarity

SN Critique Suggestion
1 Visualisation on the left is intended to show “which country is more pro-vaccine”, by computing the percentage of each response and combining them into a percentage stacked bar chart. However, for each country, it’s not easy to see the proportion of pro-vaccine and compare it among different countries. Sort countries by proportion of pro-vaccine responses in descending order.
2 Moreover, scaling each bar into the same height is not clear enough to compare the difference between pro-vaccine responses and anti-vaccine responses. Place neutral responses at centre 0, negative value showing proportion of disagree responses and positive value showing proportion of agree responses.
3 It’s hard to distinguish bars having the same length. Add value label on each bar.
4 Legend is not very clear, i.e. what does 2, 3, 4 represent specifically? Change 2, 3, 4 to 2 – Agree, 3 – I don’t know, 4 – Disagree respectively.
5 Visualisation on the right is generally straightforward and clear. However, the simple percentage does not reflect statistical measures, how much you can expect your service results to reflect the view from the overall population. For example, if we have a high proportion of strongly agree responses with a wide margin of error, then this survey result is not very reliable. Calculate confidence interval for each value to get a more comprehensive view of the survey results.
6 Moreover, this visualisation is not convincing enough in a way that only high level view of the survey results is presented and insights from different angles are not revealed. For example, by examining deeper into gender level, males are generally more willing to get vaccinated than females respondents. Explore survey results from various perspectives, such as gender and employment status, and the relation between other survey questions.

2.2 Aesthetics

SN Critique Suggestion
1 It’s redundant and distractive to use five different colours for each response in the left visualisation, making it hard to view the survey results. | Use the same colour for the same group (1.agree 2.neutral 3. disagree) with different hue level. For example, dark green for strongly agree, light green for agree. Use the same colour for the same group (1.agree 2.neutral 3. disagree) with different hue level. For example, dark green for strongly agree, light green for agree.
2 The x-axis of two visualisations are not consistent, the first plot has no decimal place whereas the second plot has 2 decimal places. Make them consistent.
3 Generally good axis marks in twenties and grid lines to facilitate easy readings, clear use of fonts, font sizes and layout with very straightforward titles. Follow and format to ensure so.

3. Proposed Design

3.1 Sketch

Figure 2: sketch of proposed design

3.2 Advantages of Proposed Design

The first visualisation:
1. Clearly show which country is more pro-vaccine, by sorting rows according to % of pro-vaccine.
2. Easier to detect difference between two types of responses and difference between countries, with postive and negative x-axis representing pro-vaccine and anti-vaccine responses respectively.

The second visualisation:
- Clearly show how reliable the result is, by applying confidence interval.

The third visualisation:
- Help audiences better understand the public willingness on Covid-19 vaccination and the potential reasons why the public agrees or disagrees to getting vaccinated, by looking into the result of survey’s questions vac2.1(worry about getting COVID) and vac2.2(worry about potential side effect of vaccine), and plotting their relation wtih trend lines.

Other comments:
- By applying tooltip, the fourth and fifth visualisation provide audiences with a more comprehensive understanding of the survey results on vaccine willingness at the country level. Similarly, being able to filter via gender and employment status gives insights on the different behaviours within each group.

4. Data Visualisation Step-by-Step

4.1 Data Preparation

  1. Import australia.csv, canada.csv, denmark.csv, finland.csv, france.csv, germany.csv, italy.csv, japan.csv, netherlands.csv, norway.csv, singapore.csv, south-korea.csv, sweden.csv, united-kingdom.csv files into tableau.  
  2. Double click on New Union at the bottom left of Files and drag all csv files into Specific. Figure 3: union imported files
  3. Click on the triangle button on the top right and custom split variable vac_1 to obtain the score value for each response category (e.g. 1 - Strongly agree -> 1). Figure 4: select custom split Figure 5: custom split
  4. Edit aliases for new column created (i.e. vac_1 - Split 1), 1 -> 1 - Strongly, 2 -> 2 - Agree, 3 -> 3 - I don’t know, 4 -> 4 - Disagree, 5 -> 5 - Strongly disagree. Figure 6: custom split Figure 7: edit aliases before Figure 8: edit aliases after
  5. Rename new column as vac1_score.
  6. Click on “=Abc” symbol and change its data type to Number(whole). Figure 10: change column’s data type
  7. Rename Table Name as Country.

4.2 Creating Visualisation

4.2.1 Visualisation 1

  1. Open a new worksheet.

  2. Click on Analysis -> Create Calculated Field.
    Figure 12: create a calculated field

  3. Create 7 calculated fields as follow:

    • Number of Records - Vac1
      Figure 13: calculate number of records that are not null
    • Total Records - Vac1
      Figure 14: calculate total records
    • Count Negative
      Figure 15: Count Negative
    • Total Count Negative
      Figure 16: calculate total count negative
    • Percentage - Vac1
      Figure 17: calculate percentage of records
    • Gantt Start
      Figure 18: Gantt Start
    • Gantt Percentage
      Figure 19: Gantt Percentage

  4. Drag Country to Rows, Grantt Percentage to Columns, vac1_score to Detail and Color

  5. Gantt Percentage, click on the triangle button -> Compute Using -> vac1_score
    Figure 21: change gantt percentage’s compute using

  6. On the legend panel, right click on Null, exclude null records.
    Figure 22: exclude null

  7. On the Marks panel, right click on the triangle button of vac1_score (either color or detail), manually sort the vac1_score, in a order of strongly disagree -> disagree -> I don’t know -> agree -> strongly agree.

  1. Change chart type from Automatic to Gantt Bar
    Figure 28: change to gantt bar

  2. Drag Percentage - Vac1 to Size and Label.

  3. Adjust label’s alignment.
    Figure 30: adjust label alignment

  4. Change label’s format from Automatic to Percentage.

  1. Similarly, right click on the x-axis and choose format, change axis’s value format to Percentage and add tick marks.
  1. Change color of the gantt bar. What it displays here is already the color I wanted. Red color represents anti-vaccine responses and green color represents pro-vaccine responses. Also, darker color implies more extreme responses (e.g. strongly agree and strongly disagree).
    Figure 36: change color of gantt bar
  2. Sort Country by % of pro-vaccine responses.
    Figure 37: sort gantt bar
  3. Right click on x-axis Edit Axis, change x-axis’s range to -100% to 100% and change its title to % of Total.
  1. Drag gender and employment_status to Filters.
  1. Add pro-vaccine and anti-vaccine annotation on the chart for better clarity.

4.2.2 Visualisation 2

  1. Open a new worksheet.
  2. Create 11 calculated fields as follow:
    • Number of Strongly Agree - Vac1
      Figure 47: Number of Strongly Agree - Vac1
    • Prop of strongly agree - vac1
      Figure 48: Prop of strongly agree - vac1
    • Z_95%
      Figure 49: Z_95%
    • Z_99%
      Figure 50: Z_99%
    • Prop-Standard Error
      Figure 51: Prop-SE
    • Prop_Margin of Error 95%
      Figure 52: Prop_Margin of Error 95%
    • Prop_Margin of Error 99%
      Figure 53: *Prop_Margin of Error 99%
    • Prop_Confidence Interval 95%
      Figure 54: Prop_Lower Limit 95%
      Figure 55: Prop_Upper Limit 95%
    • Prop_Confidence Interval 99%
      Figure 56: Prop_Lower Limit 99%
      Figure 57: Prop_Upper Limit 99%
  3. Drag Country to Rows and Prop of strongly agree - vac1 to Columns
  4. Drag Measure Values to the top of x-axis.   Figure 58: drag measure values
  5. In the Filters panel, click on Measure Names -> Edit Filter -> select 95% and 99% upper and lower limits.
  1. Change 95% and 99% confidence interval representation to Line, click on the path and choose the last Line Type.
  1. Drag Measure Names to path under Measure Values section.
    Figure 63: drag measure values to path
  2. Click on Measure Names (Detail) under Measure Values section and Prop of strongly agree - vac1 section, change it to Color.
  1. Change columns order, put Measure Values before Prop of strongly agree - vac1 and right click on the top x-axis -> synchronize axis.  
  1. Change the Measure Names’ order in the color legend, by adjusting in the Measure Values section.
  1. Hide the top x-axis, by unticking Show Header.
    Figure 72: hide header
  2. Sort rows according to % of strongly agree.
    Figure 73: sort data
  3. Change the color for 95% and 99% confidence interval according to your own preference, change axis’s number format and add tick marks. Detailed procedures are described in the visualisation 1.
  4. Final look of visualisation 2.   Figure 74: visualisation 2 final look

4.2.3 Visualisation 3

  1. Open a new worksheet.
  2. Create 8 calculated fields as follow:
  1. Drag Prop of strongly agree - vac1 to Columns, Prop of strongly agree - vac2.1 and Prop of strongly agree - vac2.2 to Rows, also drag Country to Detail under All section.
  2. Drag Measure Names to Filters and only tick Prop of strongly agree - vac2.1 and Prop of strongly agree - vac2.2.
    Figure 84: mn to filter
  3. Drag Measure Names to Color under All section.
  4. Right click on the pane to add two trendlines.
    Figure 86: add trendline
  5. Drag Country to Label and highlist points wiht minimum % of strongly agree 2.1/2.2 and maximum % of strongly agree 2.1/2.2.
  1. Change axis’s number format and add tick marks. Detailed procedures are described in the visualisation 1.
  2. Final look of visualisation 3.   Figure 89: final look v3

4.2.4 Visualisation 4

  1. Open a new worksheet.
  2. Drag Country to Rows and Prop of strongly agree - vac2.1 to Columns, drag another Prop of strongly agree - vac2.1 to the top of x-axis and synchronize axis.
    Figure 90: dual axis
  3. Change one of the variable representation to Bar.
    Figure 91: convert to bar
  4. Adjust size and color of Bar and Circle.
    Figure 92: size of bar
    Figure 93: color of bar
  5. Drag Country to Label under Prop of strongly agree - vac2.1 (Circle) section and click on Label to just its Alignment.
    Figure 96: add label
  6. Change the label format to Percentage.
  1. Sort rows according to % of strongly agree - vac2.1, change axis’s number format and range, add tick marks. Detailed procedures are described in the visualisation 1.
  2. Final look of visualisation 4.   Figure 99: final look v4

4.2.5 Visualisation 5

Same procedure as visualisation 4, except using Prop of strongly agree - vac2.2 instead of Prop of strongly agree - vac2.1. Final look of visualisation 5:
Figure 100: final look v5

4.2.6 Visualisation 6

  1. Open a new worksheet.
  2. Drag employment_status to Columns, and two vac1_score to Rows, modify both of them to measure Count.
    Figure 101: measure count
  3. Right click on the bottom y axis to make the plot into dual axis.
    Figure 102: dual axis v6
  4. Exclude Null employment status.
  5. Change one of the vac1_score representation to Line.
    Figure 104: change to line v5
  6. Drag vac1_score to Label under the circle representation and modify it into count.
    Figure 105: add label v5
  7. Apply Quick Table Calculation - Percent of Total to all three CNT(vac1_score).
    Figure 106: % of total
  8. Keep only one y-axis, change the name of y-axis. Detailed procedures are described in the previous visualisation.
  9. Final look of visualisation 6.
    Figure 107: final look v6

4.2.7 Visualisation 7

Same procedure as visualisation 4, except using gender instead of employment_status. Final look of visualisation 7:
Figure 108: final look v7

Add Tooltip on Visualisation 1 & 2

  1. In the visualisation 1, click Tooltip under the Marks, select the sheets you want to add.
  1. In the visualisation 2, click Tooltip under the Prop of strongly agree - vac1 section, modify the tooltip according to your needs.

4.3 Creating Dashboard

  1. Drag sheets you want to present into the canvas.
    Figure 115: dashboard 1
  2. Add title and subtitles (using Text object).
  1. Change subtitles’ background color.
    Figure 118: dashboard 4
  2. Add filters Gender & Employment Status to the dashboard.
    Figure 119: dashboard 5
  3. Chnage filters’ form of representation.
    Figure 120: dashboard 6
  4. Use Blank to adjust dashboard’s presentation.
    Figure 121: dashboard 7
  5. Final look:
    Figure 122: dashboard 8

5. Major Observations

Figure 123: ob 1
Figure 124: ob 2 Figure 124: ob 7 1. UK respondents showed the highest willingess to get vaccinated, regardless of female or male. This is consistent with the result of question Vac2.2 with respect to potential side effect. As shown in the relationship chart, UK respondents are also least worried about the side effect of COVID vaccine.
2. Males are generally more pro-vaccine than Females.

Figure 125: ob 3 3. Japan’s result is very interesting. Japanese respondents show low willingess in getting vaccine overall, but they also express the highest worries of getting COVID compared to other countries. 38% of Japanese respondents worry about getting COVID whereas less than 38%, 35% of respondents worry about the potential side effects of COVID vaccine, implying getting COVID is more of their concern.

Figure 126: ob 4 4. The survey results is overall reliable, with a narrow confidence interval for all countries. Netherlands has a relatively larger CI range, but it’s acceptable.

Figure 127: ob - K Figure 128: ob - J Figure 129: ob - S 5. By looking into the results of more developed economies in Asia, Japan, Singapore and South korea, we can see that full-time employed respondents are more willing to get vaccination in all three countries, compared to other employment status.

Figure 130: ob - F
6. French respondents show the least interest in getting vaccinated, which could possibly be explained by the fear of potential side effect of COVID vaccine as almost half of respondents(44%) strongly worry about the side effect.